Safety Score

[BillC]

On my 4th trip today got a "Unsafe Following" which makes no sense if UF only kicks in at 50 mph. I was using AP anytime I was driving 50 mph or higher. Despite what TSLA says, I think Safety Score is tagging UF below 50 mph -- weird

 

[Daniel in SD]

My hypothesis is that you'd be more worried about paying for the body work when attempting to put a 2-ton vehicle on its front grille.

My proposed solution to the regen while going downhill problem:

Carolina squat!

 

[kbM3]

Tesla defines hard braking as >0.3g like 6.7mph decrease of speed in one second.

If that's on flat ground, then if you instead are going down a steep hill you're going to have to decelerate higher than 0.3g (vector sum) in order to stop at 6.7mph per second.

Ever stopped suddenly at the bottom of a steep hill, your stuff all slides forward and you are forced against the seatbelt. Stop at the same rate going up a hill and stuff barely moves. Braking (backward accelerometer reading) is much higher going down a hill because of gravity trying to speed you up. That's assuming they are measuring by accelerometer and not rate of decrease in speed.

Except that there have been multiple people complaining about being dinged for regen braking on downhill. The regen is programmed for a fixed deceleration but when you add the deceleration to the acceleration due to gravity it you could exceed the hard braking threshold.
My hypothesis is that if you were stick the car on its nose the system would register 1g of braking force because it's using the accelerometer to measure both lateral (turning) and longitudinal (braking) force.

0.3 g is 6.7 mph / s by definition of the term. Also Tesla specifically stated backwards.

The reason things might go forward in the car on a steep hill under gentle braking is gravity.

The combined acceleration vector of gravity and vehicle acceleration applies even when braking on level ground. It’s the combined vector of 1 g down from gravity and 0.3g backwards (for example) for braking.

I would assume Tesla is not factoring in gravity in the calculation. It should only look at the the change in velocity of the vehicle, not the vector of the force acting on the passengers.

They should only be measuring deceleration and triggering hard braking count only when decel exceed 6.7 mph / s.

I would assume those who thought they triggered downhill, braked harder than they thought (probably at the bottom) or actually triggered it going uphill.

I would assume longitudinal deceleration is calculated by just differentiating velocity.

Paging @mongo to weigh in on acceleration calculation/measurement and the force of gravity (pardon the pun

 

[BillC]

Anyone know if the 7 consecutive days of good driving scores is rolling? If on day 3 you hit a run-away shopping cart at Publix does your Safety Score keep averaging until you finally meet the required score or is it "game over"?

 

[mhan00]

0.3 g is 6.7 mph / s by definition of the term. Also Tesla specifically stated backwards.

The reason things might go forward in the car on a steep hill under gentle braking is gravity.

The combined acceleration vector of gravity and vehicle acceleration applies even when braking on level ground. It’s the combined vector of 1 g down from gravity and 0.3g backwards (for example) for braking.

I would assume Tesla is not factoring in gravity in the calculation. It should only look at the the change in velocity of the vehicle, not the vector of the force acting on the passengers.

They should only be measuring deceleration and triggering hard braking count only when decel exceed 6.7 mph / s.

I would assume those who thought they triggered downhill, braked harder than they thought (probably at the bottom) or actually triggered it going uphill.

Paging @mongo to weigh in?

I literally never hit the brake in my drives. Got dinged for a hard brake event on my downhill portion to get to the stores I needed for my errands. Raised my score for the day from 95 to 97 on my drive back that was going uphill. I’ve gotten pretty good at never needing to hit the brakes at all on a good percentage of my drives and using pure regen. That’s most of my AP disengagement’s too: I disengage when approaching stopped or slowing traffic not because I’m worried about AP running into them (though of course that’s something to keep in mind for stopped traffic), but because I think AP uses some brakes and not pure regen.

 

[Daniel in SD]

I would assume longitudinal deceleration is calculated by just differentiating velocity.

I assume they're using the accelerometer and are simply measuring the longitudinal force.
I'm glad we agree on what the question is though.

In their giant curve fit that they used to create this metric I bet it's slightly better to use longitudinal force since braking distance is increased when going downhill.

 

[DreamBurn]

I just went to the app store and searched Tesla. It will show an update. It's not automatic updates. It'll be the same for Android. I had to manually search it last time there was a refresh too.

Wow... you're right... the automatic update just updated me to 4.0.something... I had to search the Apple App Store for Tesla and then it showed me the update to 4.1.0. THANKS!

 

[EVNow]

That's more likely because California state law (Prop. 103) prohibits insurance companies from using telematics to set premiums. They must be calculated based on safety record, length of driving experience, and miles driven annually.

Thats explains it.

So, Tesla never bothered to check the safety rating for accuracy. They will have to vastly improve it before its usable.

 

[Terminator857]

Appreciate all you thirsty people who rushed out to set this up / "push the button". I have FSD but am in no particular hurry to push this button. Maybe in a couple months or so, for me.

Thanks again for all you alpha -2 testers

Everyone who gets the beta: You'll need to talk about how your mind is blown, quantum leap, best thing since sliced bread, better than a disney ride and how everyone is missing out big time!

 

[fiveten]

I worked on trying to improve my aggressive turning metric today. More miles on autopilot doesn’t seem to effect it. Driving around in a round about for 2 minutes straight didn’t seem to improve it either. However slowly driving around my neighborhood taking as many turns as I could for about a mile and a half drive took me from 5% for the day to 2.3% (I had 39 total miles so far today, the majority on autopilot)

 

[fiveten]

I worked on trying to improve my aggressive turning metric today. More miles on autopilot doesn’t seem to effect it. Driving around in a round about for 2 minutes straight didn’t seem to improve it either. However slowly driving around my neighborhood taking as many turns as I could for about a mile and a half drive took me from 5% for the day to 2.3% (I had 39 total miles so far today, the majority on autopilot)

Im at 2.7% at 146 total miles in my Y, gonna drive around my neighbor for 2 miles taking turns and see if i can halve that.

 

[gcsmoore1]

Model 3 here with software update 2021.32.22
Went through the procedure to authorize the trial period.
Car has no detectable change in behavior.
Can someone tell me how to view my driving report please?
Check my updated app, and can't find it.

 

[EVNow]

I assume they're using the accelerometer and are simply measuring the longitudinal force.

My assumption is whoever coded this for safety rating has no idea how the acceleration is measured - they are just using something that was already available (one of hundreds that the car probably has available on the bus).

 

[Dave storer]

Model 3 here with software update 2021.32.22
Went through the procedure to authorize the trial period.
Car has no detectable change in behavior.
Can someone tell me how to view my driving report please?
Check my updated app, and can't find it.

In order to see your safety score, you must have the app 4.1.0, The app on Google Play is 4.0.2. Not sure when Play store will get the right version.

 

[Aerodyna]

Tip of the day: While driving or after trip started, if you reboot your system that trip won’t be recorded in your safety score calculation!

I tried this today and it worked for me. While driving manually at 20mph, i got a forward collision warning for a car stopped in front of me. I was closely watching traffic and getting ready to slow down but system reacted before me and gave an so unnecessary warning. So after that i pulled over and reboot my system by pushing two scroll buttons. Then after trip ended i checked my safety score and realized system didn’t recognized my trip Hope this helps.

 

[fiveten]

Im at 2.7% at 146 total miles in my Y, gonna drive around my neighbor for 2 miles taking turns and see if i can halve that.

After a 2.9 mile drive, I’m now at 2.2%

 

[Daniel in SD]

My assumption is whoever coded this for safety rating has no idea how the acceleration is measured - they are just using something that was already available (one of hundreds that the car probably has available on the bus).

I agree. I'm just saying that it might be slightly more predictive of collision rate than the change in speed. People who don't increase their braking distances on hill might have more collisions.

So, Tesla never bothered to check the safety rating for accuracy. They will have to vastly improve it before its usable.

They definitely checked it for accuracy, how else would you calculate the coefficients?
I assume that their model is perfectly fit to the data (look how many digits there are!). I'm sure there could be a much better model, it would be interesting if they told us how accurate this one is. It would be fun if they had a contest where they released a billion miles of data and people could write their own prediction model (I remember Netflix doing this early on for customer movie ratings in order to make suggestions).

 

[kbM3]

I agree. I'm just saying that it might be slightly more predictive of collision rate than the change in speed. People who don't increase their braking distances on hill might have more collisions.

They definitely checked it for accuracy, how else would you calculate the coefficients?
I assume that their model is perfectly fit to the data (look how many digits there are!). I'm sure there could be a much better model, it would be interesting if they told us how accurate this one is. It would be fun if they had a contest where they released a billion miles of data and people could write their own prediction model (I remember Netflix doing this early on for customer movie ratings in order to make suggestions).

Why would it be more predictive of collision? According to that theory, it would be unsafe to drive down a steep hill at a constant velocity and it would be safer to accelerate down a steep hill.

 

[kbM3]

I assume they're using the accelerometer and are simply measuring the longitudinal force.
I'm glad we agree on what the question is though.

In their giant curve fit that they used to create this metric I bet it's slightly better to use longitudinal force since braking distance is increased when going downhill.

That would definitely be an error on Tesla’s part as they specifically said ”This is the same as a decrease in the vehicle’s speed larger than 6.7 mph, in one second”.

And the vehicle would think you’re hard braking, if the car were stopped on a 45 deg. decline.

I find it incredibly difficult to believe Tesla would make that poor of an engineering choice and describe it incorrectly in the safety report.

I’m betting people are slowing their vehicle at more than that at some other point on the drive, than on a substantial downhill.

 

[KArnold]

After a 2.9 mile drive, I’m now at 2.2%

Looks like "hard braking" is weighted more heavily than "aggressive turning" - see the Safety Score Indicator to play "what if" games.

The former is defined as "proportion of braking time spent braking with extreme force". If making many easy turns in your neighborhood reduces that score, I'll bet the same may be true for braking as well. Do you recall those before/after stats?